Show me the data!

The TL;DR summary: In February 2017, when Elsevier were accused of selling one paid-for hybrid open access article, at first they sowed doubt about it, then three days later admitted it to be true. In their admission they state that it is the only wrongly paywalled open access article “affected” at their websites. They have apparently checked their “system” and say there are no more wrongly paywalled articles at Elsevier websites. Their statement is however demonstrably incorrect as I show in this post. There are more paywalled open access articles at Elsevier, and by their own admission Elsevier don’t seem to know about them. This is professional incompetence. HybridOA as whole, not just at Elsevier, is not reliable. We should not use hybridOA services if we actually want reliable open access in perpetuity. 

Some Background

Elsevier and other legacy publishers, have an unfortunate track record of paywalling articles that research funders have paid-for to be openly accessible. Elsevier in particular are notable for their consistency in this regard: they have been caught by diligent readers, selling access to “open access” articles in 2014, 2015, 2016 and 2017. They are so notable for this practice that it is on their Wikipedia page amongst a long section of other controversies.

Responding to the March 2014 scandal, Elsevier claimed it was merely “bumps on the road to open access”. Shortly before Christmas (a great time to bury bad news) on the 22nd December 2014 Elsevier quietly disclosed that they had refunded “about $70,000” to readers across the world who had purchased access to articles that should have been free to access at Elsevier. Naturally, as this self-disclosure came just before Christmas it went largely unnoticed and unreported.    

Now we fast-forward to February 2017. As an academic, fed-up with abysmally poor publishing services, I have decided to systematically run heuristic checks on articles that Wellcome Trust has paid-for to be open access. The Wellcome Trust and other research funders have started to publish detailed reports on which articles they have paid-for to be open access, with which publishers, and how much they paid publishers to make each article open access in perpetuity. I have used these public records supplied by Wellcome Trust to cross-check articles that are supposed to be open access against what is actually open access at publisher websites.

From this exercise so far I have found at least seven Wellcome Trust paid-for “open access” articles that are currently paywalled, and for sale to readers for up to $58.80 per article, at major academic publisher websites. I have publicly given details of one of these articles so far: paywalled at Elsevier at a journal called Mitochondrion.

The first response: sow doubt on the claim

Initially, Elsevier responded to my observation on a public mailing list:

Ross – we are looking into this, but an initial glance at the article suggests that the author first decided to publish gold OA and then decided against it.  We’ll have a little look to make sure all the metadata is correct.

— Dr Alicia Wise, Director of Access and Policy, Elsevier | Feb 15th 2017 09:21:44 GMT (source)

Note that Elsevier’s initial response suggests that I might have been incorrect in my assertion that this article should be open access and should NOT be on sale to readers.

I am a scientist. I make statements based upon evidence. I am no fool. But Elsevier seem to think I am.

The second response: admit to one error but clearly deny any further extent

Three days later Elsevier admitted that I was correct – the Wellcome Trust had in fact already paid Elsevier £2,168.08 many years ago to make this article open access, it should have been open access at their website, and should NOT be on sale to anyone.

Hi Ross, this has taken a little time to bottom out but indeed the article should be OA. In 2013 you may recall we initiated an exercise to add/correct and generally clean up OA license information on articles.  This article was part of that exercise, and a little unusual: we had invoiced the author, and then the institution wanted to receive the invoice but in a different currency, so the original  article invoice was unpaid during the cleaning exercise.

Our normal practice is to publish articles OA, but to move them back behind the firewall if after a grace period the invoice remains unpaid.  So it was picked up as an article with an unpaid APC and moved behind the firewall.  We’ve gone through the system, this is the only article affected. There’s nothing malicious about what happened here, and we’ll be reimbursing the institution’s APC, and your PPV charge.  We will also run a check to see if anyone else paid PPV for this article and if so reimburse them as well.

— Dr Alicia Wise, Director of Access and Policy, Elsevier | Feb 18 2017 00:03:24 GMT (source)

Within the above statement from Elsevier there is something much more astonishing. Elsevier clearly wants us all to “move along, nothing to see here” by stating that the Mitochondrion article is “the only article affected” by paywalls that shouldn’t be.

But there are more paywalled open access articles at Elsevier journals

Until now I haven’t disclosed publicly that there is at least one other paid-for “open access” Wellcome Trust article that is currently paywalled at an Elsevier journal, despite having been handsomely paid-for to be made open access in perpetuity. It’s at a rather well-known Elsevier journal too; The Lancet. According to data Wellcome Trust has made available, in 2012 they paid Elsevier £5,280.00 to make the article in the screenshot below, open access in perpetuity. In 2017 I find it is on sale at ScienceDirect for $35.95 + VAT. I bought access to it just to verify that it is actually on sale and indeed it is.

The screenshot below is my receipt documenting my purchase of the above “open access” article from The Lancet:

Now let’s return to Elsevier’s most recent statement and you’ll see why I’m shocked:

“We’ve gone through the system, this [the Mitochondrion article] is the only article affected.”

I have just provided evidence (The Lancet article) that there are more paid-for “open access” articles at Elsevier that are paywalled and have been sold to readers. As my receipt demonstrates, my purchase of this Wellcome Trust paid-for “open access” article post-dates the above statement by Elsevier. Therefore Elsevier’s statement would appear to be incorrect.

So we now have evidence of two further important things to note:

  1. Elsevier’s entire system for handling hybrid open access is broken
  2. Elsevier are evidently incapable of accurate self-assessment

Elsevier say they have gone through their “system” and that there are no other open access articles “affected” by paywalls. But I have just provided clear evidence to the contrary.

To be clear I’m not accusing Elsevier of knowingly lying about this. I am going to charitably interpret this as an unknowingly false statement – they believe it to be true from what their faulty “system” tells them, but in reality it is not true, other articles are affected. It is abundantly evident to me that Elsevier’s internal systems and processes for tracking invoices for APC payments, handling, and marking hybrid open access articles simply aren’t robust. I believe that this forms part of a body of mounting evidence that Elsevier is professionally incompetent at hybrid open access publishing and that this incompetence financially benefits them through pay-per-view payments & re-use licensing to the detriment of the scholarly poor (those outside paywalls). I don’t allege malicious intent. The point is that regardless of intent, this incompetency is demonstrably generating profits for them – that is undeniable and that it is causing real harm (financial loss to readers who have paid for articles that should not be for sale).

What is the true extent of this problem at Elsevier in 2017? Why does it matter?

I have to admit I don’t have a definite answer as to the extent, millions of articles are published every year, I can’t check them all myself. But one thing I do know and can now evidence, is that clearly Elsevier don’t know either! I think a full independent audit with forensic accounting is urgently needed. Why? As I mentioned earlier, back in 2014 Elsevier self-disclosed that they had charged readers $70,000 to access articles that should have been “open access”. Since 2014 the rate of publication of hybrid open access articles has increased so I would expect if a competent independent audit were done of purchases of ‘mistakenly’ paywalled articles between 2015 and now, the sum to be reimbursed might well be in excess of $100,000. Given Elsevier doesn’t know that some of its paid-for hybrid open access articles are on sale, can we trust their self-reported audit of such incidences prior to 2014? I think not, hence a thorough independent audit is needed.

The scale of this could be very big. I have only partially cross-checked the accessibility of hybrid open access articles from just one research funder at Elsevier. According to CrossRef data there are at least 14,155 different research funders. The charge to make articles hybrid open access at Elsevier is typically around $3000 but can be much more as we have seen from The Lancet article which was £5,280 to make open access. When found to be in error, Elsevier typically refund the APC-paid, so therefore if just one “open access” article per research funder is mistakenly paywalled, it would equate to a refund of approximately $42,465,000 USD. This amount is still chump change to Elsevier: a company that in 2013 posted an annual adjusted operating profit of £826 million on an annual turnover of £2.126 billion.

Nor in terms of scale is this scandal confined to just Elsevier. There are paywalled “open access” articles for sale at Emerald Publishing Group (exonerated 2017-02-27), Cambridge University Press and Wolters Kluwer / LWW to name but a few. I have evidence, which I have already informed Wellcome Trust of. I don’t make these claims lightly.

I stress however that this scandal is confined to hybrid open access publishing.

Only fully open access publishers are reliable

How many times have PLOS accidentally paywalled an article? [0]

How many times have PeerJ accidentally paywalled an article? [0]

Full open access publishers such as PeerJ, PLOS, MDPI, Hindawi, Copernicus, Pensoft, F1000Research, and Ubiquity Press to name but a few, have an open access compliance “error” rate of absolutely ZERO, out of hundreds of thousands of articles published so far.

In some respects these are facile statements to make about full open access publishers. But that is my point. It is known that PLOS and PeerJ cannot paywall and ‘mistakenly sell’ articles to readers because their systems are open by default and do not have paywalls. At fully open access publishers, all articles are open and thus when an APC is paid or a fee waiver awarded, there is never any failure to make articles available in a manner that is both freely accessible to readers (not paywalled) and licensed under an open access definition compatible Creative Commons licence that Wellcome Trust requires.

If research funders and institutions really want open access publication, under open access definition compliant licensing, they should use full open access publishers.

Luckily not all research funders allow their money to be used to pay for hybrid open access charges. This is sensible policy in my opinion, given the evidently incompetent open access service provided by many, not just Elsevier.

Conclusions: what can funders & APC-paying institutions do about this going forwards?

As soon as reasonably possible, stop paying for hybrid open access. It is not reliable. If funders and institutions really want open access, use preprint servers and full open access journals. Hybrid open access is demonstrably unreliable, vastly over-priced, and does not help towards long term goals of open access as it helps keep subscription-access journals going and takes good articles away from full open access journals. It is time to end the experiment.

TL;DR Elsevier are selling access to open access articles again

I saw this humorous tweet today:

This tweet references the fact that Elsevier have been caught selling access to paid-for “open access” articles in 2014, 2015, and 2016.

After the 2014 scandal they self-reported an internal audit (with no external or independent oversight) and apparently refunded a total of “about $70,000” to readers who had ‘mistakenly’ been allowed to purchase articles that should have been open access. There was no independent audit of Elsevier’s systems – this was all self-reported. Can you imagine allowing Volkswagen to self-report diesel emissions? Not a good idea!

I do not trust Elsevier to self-report on the scale of these events – I really really think governments should step-in here and external, independent auditors should be given access to their internal systems to determine independently the scale of these “mistakes” that defraud readers to the financial benefit of Elsevier.

Now as you know recently I’ve been looking at 2-years worth of Wellcome Trust open access APC payment data.

This post is just to let the world know that in 2017 Elsevier are doing it again. They are selling articles that have been paid-for to be open access.

According to public data available on Figshare, Wellcome Trust paid Elsevier £2,168.08 to make open access an article entitled:

A multi-center comparison of diagnostic methods for the biochemical evaluation of suspected mitochondrial disorders paid to be open access on behalf of WT Grant number: 096919, Professor DM Turnbull, Newcastle University.

Today (14th February 2017), from Elsevier’s ScienceDirect website, I found that this article was behind a paywall. To be able to access this at home (outside the paywall), Elsevier charged me $43.14 including tax, in return for 24 hours access to this paid-for “open access” article. I bought access to the article to prove beyond doubt that Elsevier really were selling this open access article and it wasn’t just some fleeting browser error that they could explain away.

Below is a screenshot of my emailed electronic receipt as proof, with my home address crudely redacted to protect my privacy.

If you are a journalist and would like to talk more about this case, or the history and extent of ‘paywalled open access’ please email me at:

In the mean time I shall be contacting both the Wellcome Trust and Elsevier to alert them about this issue, again.




Yesterday I published a blog post calling for ongoing monitoring of ‘hybrid’ open access articles and academic publisher services in general.

Today I want to share with you some highlights from my brief checks on 2 years worth of Wellcome Trust ‘open access’ article processing charge (APC) supported published research outputs.

Source data

Robert Kiley of the Wellcome Trust has made public official data on the APC spend on ‘open access’ articles paid for by the Wellcome Trust over at his figshare profile. This was a brilliant thing to do. Many people have made thought-provoking and brilliant analysis of this data. The data has been copied many times and I see it now in many different github repositories.

Yesterday, just to test the idea, with no real knowledge that I’d actually find anything of interest, I decided to check the DOIs of 2 years worth (2012 to 2014) of Wellcome Trust funded ‘open access’ articles. Here’s the three major things that I have discovered from this mini exercise so far:


1.) Paywalled Articles That Should Be Open Access Wellcome Trust Funded articles that are not openly accessible (updated for accuracy 2017-02-27)

According to Robert Kiley’s figshare data for 2013-2014 Wellcome Trust paid £1,194 to Emerald to make an article entitled Running a hospital patient safety campaign: a qualitative study open access at the publisher website. I followed the DOI link given and found that today this article is paywalled, and is being advertised for sale at £20 for 30 days of access by Emerald Group Publishing (screenshot below):


Sadly, I am no stranger to this kind of event. I have personally seen Elsevier, Wiley, Springer and Oxford University Press sell articles that had been paid-for specifically by funders to be open access to everyone in the world, not to be sold at the point of availability. It seems inevitable now that hybrid open access would lead to this. Paywall publishers simply can’t keep the paywalls off, even if they are paid to do so.

UPDATE 2017-02-27: Three weeks after publication of this blog post Wellcome Trust and Emerald kindly confirmed to me that no APC was actually paid for the above article (contrary to what was mistakenly stated in the figshare data). The article authors backed-out of choosing gold open access. Unfortunately, the authors did not self-archive a freely available version of this paper either, so it remains not freely accessible to those outside paywalls and thus was definitely NOT published in a manner compliant with Wellcome Trust rules and regulations that were in place at the time.


2.) Misuse of funds set aside to cover Open Access charges

I thought this was another simple case of hybrid open access being ‘mistakenly’ paywalled by the publisher but the truth is even stranger.

I found this article entitled, Mechanisms underlying cortical activity during value-guided choice at the journal Nature Neuroscience, that according to Robert Kiley’s data, the sum of £1,272.86 had been paid by Wellcome Trust to make this article open access. The plot thickens however. Nature Neuroscience doesn’t really do hybrid open access, and if it did, it would charge a lot more than that. What I have discovered here is an instance where the authors (or their institution) have mistakenly (fraudulently?) depending on how forgivingly you view this: used the Wellcome Trust Open Access fund to pay Nature Neuroscience £1,272.86 for colour figures. The article is NOT open access at the publisher website. The 2017 Wellcome Trust guidelines are absolutely clear that you cannot use Wellcome Trust Open Access money for “page charges” or “colour figure” charges. I do not know if Wellcome’s rules were so clear back in 2012 when the payment was made but this is extremely disappointing to observe. Charity funding should be better spent than on spurious publisher-invented ransoms like “colour figure charges”.

3.) Elsevier ‘open access’ articles are not accessible to all machine methods

To help check article DOIs in a simple and automated manner, I used the R package httr. The code I used is available as a github gist. The code works well for articles hosted at all publishers except one: Elsevier. Any attempt to follow DOI links with R::httr just hangs and I have to use a timeout to ensure that my script skips over such problems to proceed onto the next article.

Can Elsevier really call what they are offering ‘open access’ if it is not openly accessible by automated methods such as R::httr scripts? I don’t have time to expound upon this at length here, but I will certainly return to this particular point at a later date.


So there you have it. Super simple automated checks of just a few thousand Wellcome Trust funded ‘open access’ articles by their DOIs has revealed three rather interesting things, and supports my overall thesis that we need to continuously monitor academic publishers: not just “one-time” compliance checks.

I really do think this is the start of something very interesting. I have plans. WATCH THIS SPACE!

In a recent series of posts I’ve become fascinated with how unnecessarily fragile the scholarly communications system seems to be in 2017:

Oxford University Press have failed to preserve access to the scholarly record (23-01-2017)

Documenting the many failures across OUP journals (24-01-2017)

Comparing OUP to other publishers (25-01-2107)

As a reminder, academics literally invented the internet, I think we can and should be doing better.

We have the technology and resources available to make a robust and efficient scholarly communications system, yet from the more than $10 billion per year we spend on it every year we appear to be getting incredibly poor service from our various providers. If we take Digital Object Identifiers (DOIs) as an example – they are great in theory. If I send a colleague a DOI such as “10.1093/sysbio/syw105” in theory, in five years time, no matter who owns this journal, or what platform technology they decide to use, my colleague should be able to follow this DOI-based link to an article landing page: (as of 06-02-2017 this one happens to be broken though!).

The DOI registration agency (CrossRef) that most journals use is highly competent. I have no doubts about their technical abilities, or the support and documentation they provide to publishers. Most modern, born-digital publishers create DOIs for their journal articles with accurate metadata, and little difficulty or breakage. Yet when it comes to some publishers like Oxford University Press, I am amazed to find that more than 2.3% of their DOIs result in a 404 “Not Found” error. Collectively, research institutions, libraries, and personal subscribers across the world pay publishers like OUP a huge amount every year to provide publishing services. Why can’t they actually do the job we pay them to do?

Personally, I have now lost all faith in the ability of many legacy academic publishers to actually publish content on the internet in a robust manner. For example, when we pay them >$3000 to make an article hybrid open access, what happens? They end up putting it behind a paywall and selling it to readers, even despite the hybrid open access payment. Wiley, Elsevier, Springer and now OUP have been caught doing this to some extent. Nor are the failures confined to just subscription access journals. Remember when all of Scientific Reports and other NPG journals were down for a few days with absolutely no prior warning (mid June 2016)? I totally understand the need for publishers to upgrade and or maintain their platforms but the apparent lack of testing or forethought when they decide to fiddle with their platforms is professionally incompetent at times, and this recent problem with OUP definitely falls into this category. When GitLab (not an academic publishing services company, unfortunately) made a serious screw-up recently that caused a major outage to their services, they livestreamed on Youtube their attempts to fix the problem and had it fixed in about 12 hours. They also published full, transparent, extensive and frank reports on what happened, why it happened and how and when they fixed it. In stark contrast to this OUP have given their customers opaque reassurances in an official statement AND still haven’t fixed many of the problems even weeks after! Legacy academic publishers and other modern internet services are sadly miles apart in the levels of service and transparency they provide.

So what can we do to remedy this abominable situation?

I propose that funders, institutions, and authors need to start doing more than just “one-time” compliance checks on the way that research outputs are published. We need continuous, daily, weekly, monthly or quarterly checks on research outputs just to make sure they are still actually there and that vital links to them like DOIs actually work! Additionally, those that pay for these publisher services actually need to start checking that publishers are actually providing good service. Withhold money, or bring consequences to those publishers who provide poor service.

To this end I have created toy code for use in R to help empower authors to check that the DOIs of their own authored research outputs actually work. I have my script setup as a cronjob scheduled to check my DOIs every day. Today’s fully-automated report is below, the 401 error tells me that my letter is behind a paywall at Nature (true), unfortunately, but it’s not a 404 so it’s otherwise okay:

"Success: (200) OK",""
"Success: (200) OK",""
"Success: (200) OK",""
"Success: (200) OK",""
"Success: (200) OK",""
"Success: (200) OK",""
"Success: (200) OK",""
"Success: (200) OK",""
"Success: (200) OK",""
"Success: (200) OK",""
"Client error: (401) Unauthorized",""
"Success: (200) OK",""

This idea gets more interesting however if scaled-up to the institutional or funder-level. What would happen if Cambridge University checked the DOIs of all “their” co-authored research outputs every week? What would happen if the Wellcome Trust checked the DOIs of all their funded research outputs every month? At this scale it could be done both time and cost effectively, and it is much more likely to uncover the abundant problems that lie quietly unobserved and under-reported.

Tomorrow I will blog about what I discovered when I checked the DOIs from just 2 years worth (2012 to 2014) of Wellcome Trust funded research. I promise it’ll be interesting and it’ll demonstrate more the utility of this exercise…